Preliminary Evaluation of the

1
Preliminary Evaluation of the June 2002 Version
of CMAQ Brian Eder Shaocai Yu Robin
Dennis Jonathan Pleim Ken Schere Atmospheric
Modeling Division National Exposure Research
Laboratory Environmental Protection Agency RTP,
NC 27711 On assignment from Air Resources
Laboratory, NOAA
2
CMAQ

• ? 2002 Base Version Configuration
• - SAPRC99 gas-phase chemistry
• - BEIS 3.09 biogenic emissions
• - AE3 aerosol module

3
? Spatial Domain

• 32 km grid resolution
• 22,072 grid cells

• Temporal Domain
• - 1- 14 July, 1999

4
Species, Metrics and Networks

• O3 Hourly, max 1-hr, max 8-hr (AIRS)
• SO2 Mean weekly (CASTNet)
• SO4 Mean weekly (CASTNet), Mean 24-hr
  (IMPROVE)
• NO3 Mean weekly (CASTNet), Mean 24-hr
  (IMPROVE)
• HNO3 Mean weekly (CASTNet)
• PM2.5 Mean 24-hr (IMPROVE)
• OMC
  Mean 24-hr (IMPROVE)
• EC Mean 24-hr
  (IMPROVE)
• NH4 Mean weekly (CASTNet)

5
AIRS Aerometric Information Retrieval
System Hourly data for each of the
14 days 1,032 Stations234,384 Obs
O3
6
IMPROVE Interagency Monitoring of PROtected
Visual Environments 24-hr means for 3, 7, 10,
14 July 50 Stations 200 Observations
SO4 NO3 PM2.5 OMC EC
7
CASTNet Clean Air Status and Trends
Network Weekly means for 6 13 July 59
Stations, 59 Observations
SO2 SO4 HNO3 NH4
8
Statistics - Errors

Normalized Mean Error Root Mean Square
Error Mean Absolute Gross Error Mean
Normalized Gross Error

9
Statistics - Biases
Mean Bias Mean Normalized Bias Mean
Fractionalized Bias Normalized Mean Bias
10
Ozone
11

• Ozone
• - Hourly

CMAQ 22.0 0.68 OBS

• CMAQ Performance
• - Biases positive and
• generally small
• MNB, MNGE inflated due to
• small hourly obs.

12
? Ozone - Max 1-hr
CMAQ 28.0 0.67 OBS

• CMAQ Performance
• - Biases smaller, still positive
• - Errors smaller
• MNB, MNGE closer to
• other statistics

13
? Ozone - Max 8-Hr
CMAQ 24.7 0.71 OBS

• CMAQ Performance
• Biases positive, smaller than
• hourly, larger than max 1-hr
• - Errors generally smaller

14
Daily Boxplots
? Ozone - Hourly
Observations increase from 19 July, then
decreases from 10 -12, then increase thru 14.
Simulations increase from 16 July, decrease
on 7 and again on the 10 and 11, then increase
thru 14.
15
Daily Boxplots
? Ozone Difference - Hourly - CMAQ
- AIRS
Biases are positive on all days. Largest biases
occur on 5 and 6 July Smallest occur on 1, 10
and 11 July
16
Daily Boxplots
? Ozone - Max 1-Hr
Observations increase from 18 July, then
decreases from 10 -11, then increase thru 14.
Simulations increase from 16 July, decrease
on 7 and again on the 10 and 11, then increase
thru 14.
17
Daily Boxplots
? Ozone Differences - Max 1-Hr -
CMAQ - AIRS
Biases are again positive on all days. Largest
biases occur on 5 and 6 July Smallest occur on
1, 10 and 11 July
18
Daily Boxplots
? Ozone - Max 8-Hr
19
Daily Boxplots
? Ozone - Max 8-Hr - CMAQ -
AIRS
20
Hourly Boxplots
? Ozone

• Diurnal pattern is well
• replicated by the model
• morning minimum (11 - 12 GMT)
• afternoon maximum (20 - 21 GMT)

21
Hourly Boxplots

• Ozone
• CMAQ - AIRS

• Biases are always positive
• Biases smallest in pm when
• atmosphere is well mixed.

22
? SO4 - 24-Hr mean - IMPROVE
CMAQ Performance - Most simulations within
21 - Biases are positive and small -
Errors are also small - Results better in
eastern US
23
? SO4 - Week mean - CASTNet
CMAQ 0.46 1.03 OBS
CMAQ Performance - Most simulations within
21 - Biases still positive and small -
Errors still small - Results better in eastern
US
24

• NO3
• - 24-Hr mean
• - IMPROVE

CMAQ Performance - Biases are negative and
large - Errors are also large - Few
simulations within 21
25
? HNO3 - Week mean - CASTNet
CMAQ Performance - Most simulations within
21 - Biases positive and large - MNB, MNGE
inflated due to small observed
concentrations
26
? NH4 - Week mean - CASTNet
CMAQ Performance - Almost all simulations
within 21 - Biases are positive and
large - Errors are also large
27

• PM2.5
• - 24-Hr mean
• - IMPROVE

CMAQ Performance - Most simulations within
21 - Biases are positive and large -
Errors are also large
28
? OMC - 24-Hr mean - IMPROVE
CMAQ Performance - Majority simulations
within 21 - Biases are positive and large
- Errors are also rather large
29

• EC
• - 24-Hr mean
• - IMPROVE

CMAQ Performance - Majority simulations within
21 - Biases are small and mixed - Errors
are large
30
? SO2 - Week mean - CASTNet
CMAQ Performance - Biases are positive and
very large - Errors are also very large
31
Evaluation Summary
32
Evaluation Summary

• CMAQ
• ? varies in its ability to simulated air
  concentrations of the nine species
• - correlations range from 0.55 (EC, OMC)
  to 0.90 (SO4)
• - NME range from 25 (O3) to over 100
  (HNO3, SO2)
• ? generally over-predicts
• - biases are positive (except NO3, EC) and
  for some species large
• ? performs better for O3 and SO4 than other
  species

33
Contacts
Brian K. Eder email eder_at_hpcc.epa.gov